Thursday, September 15, 2005

Back to Woodstock - Flawed Census Tract Epi and Asbestos in Water

Another exercise in census tract epidemiology, which could have been reviewed and edited better, concerns water contamination with asbestos fibers. The “feel good” “don’t worry, be happy” abstract is common for technical reports by public health authorities, but might have been better stated for a scientific publication.

GI exposure to asbestos is a meaningful issue. Rats don’t get GI tumors from eating quite a lot of asbestos, but you need to gas rats with a thousand fibers/cc to get meaningful lung tumor yield. The rat’s a lousy model for colorectal tumors, also.

Backstory: Water service goes down. “Following flushing of the water mains ,five water samples were collected from different locations in the Woodstock water supply. Four of the samples had asbestos levels greater than 10 million fibers per liter (MFL), with the maximum equaling 304.5 MFL. Over 90% of the asbestos was chrysotile; the remainder was crocidolite (Webber and Covey,1991 ).”

It would have been nice had the NYS Public Health Department provided some asbestos in water data for other than this special circumstance. Most likely there’s none because of “don’t ask, don’t tell, don’t pursue.”

Literature review: “Cancer incidence or mortality has been examined in eight populations exposed to asbestos in drinking water in the United States,Canada , and Norway …An excess of stomach cancer, primarily among males, was observed in five study populations …,and pancreatic cancer was associated with exposure among males in one population and among females in another. A positive association was not reported for more than one study population for any other cancer site.”

BrooklynDodger previously blogged, more than once, that census tract epidemiology aimed at a specific source or agent is a lose-lose scientific proposition. Absence of association is expected because these studies are likely powerless, and associations found are unlikely due to the target exposure, for the same reason.

Here, results are standardized incidence rates, adding shakiness. This blocks comparing overall mortality, to see if there is a “healthy community effect” similar to the healthy worker effect, an artifact of the method. All cancer incidence was 0.91 for the men, 0.96 for the women. Were this a big enough study to find anything, proportional cancer incidence ratios might be a more reliable indicator of excess.

Total GI cancers had Upper Confidence Intervals of 1.26 for the women and 1.63 for the men. Nevertheless, the investigators stumbled on a 3-fold significant excess of pancreatic cancer among men, with a whopping UCI. They Houdini this finding with a totally powerless residence duration response analysis for total GI [not in excess anyway] and total respiratory cancer.

No mesotheliomas were observed. The investigators might have given BrooklynDodger an expected or a confidence interval to see what level of risk had been ruled out.

What frosts the Dodger is that the editors let this line in the abstract: “This association may be related to factors other than asbestos exposure such as occupation and lifestyle or to chance.Talk about conveying negative information [i.e., stuff that makes you know less after you have read it.]

Bottom line, this study was inadequate to assess the association between drinking water and cancer in Woodstock, or asbestos in drinking water. Nevertheless, there is equivocal evidence for an association of being in this cohort and increased risk of pancreatic cancer.


Environmental Research 98 (2005) 224–232
Cancer incidence and asbestos in drinking water,Town of Woodstock,New York,1980–1998

Marilyn L. Brownea,_,Deepa Varadarajulub,Elizabeth L. Lewis-Michla, Edward F. Fitzgeralda a

New York State Department of Health, Bureau of Environmental and Occupational Epidemiology, 547 River St. Rm. 200, Troy, NY 12180, USA bNew York State Department of Health, Bureau of Chronic Disease Epidemiology and Surveillance, Albany, NY 12237,

Abstract Late in 1985,asbestos contamination was discovered in the public water supply of the Town of Woodstock, Ulster County ,New York. Contamination resulted from asbestos–cement pipes installed in the town water system in the mid to late 1950s and the corrosiveness of the local water. The New York State (NYS) Department of Health established the Woodstock Asbestos Exposure Registry (WAER) in 1986 to monitor rates of cancer among individuals who lived on the water supply between 1960 and 1985. Demographic ,health, and residential information were collected on 2936 registrants. The follow-up period for observation of cancer was 1980–1998, consistent with the expected lag of 20–30+ years for development of asbestos-related cancers. The NYS Cancer Registry was used to ascertain cancer diagnoses. Standardized incidence ratios (SIRs) for gastrointestinal, respiratory,and total cancers were all approximately 1.00 or less and all 95% confidence intervals (CIs) included 1.00. For individual types of the gastrointestinal cancers ,only the SIR for pancreatic cancer was marginally statistically significant at 2.19 (95% CI=1.00–4.16), based on a total of nine observed cases. The excess in pancreatic cancer occurred primarily among men (SIR=3.08; 95% CI=1.13–6.70) and was only slightly elevated among women (SIR=1.39; 95% CI=0.29–4.06). This association may be related to factors other than asbestos exposure such as occupation and lifestyle or to chance. No cases of mesothelioma were observed among WAER participants. There was no increase in incidence by latency or duration of residence on the water supply, but the ability to detect these trends is limited by small numbers and unknown dates of initial exposure. The general pattern of results did not demonstrate a likely link between exposure to asbestos in drinking water and cancer occurrence among participants in the WAER.

No comments: